What's New in RapidMiner Radoop 8.1.1?

Released: March 08th, 2018

The following improvements are part of RapidMiner Radoop 8.1.1.

Enhancements

  • Changed Spark Resource Allocation to default to Dynamic Resource Allocation
  • Number of attempts for all Spark jobs that is submitted by Radoop is decreased to 1. If you want to override this behaviour, add spark.yarn.maxAppAttempts property with a larger value than 1 to Advanced Spark Parameters
  • Added a warning for MapR connections on Windows, when MAPR_HOME environment variable is not set
  • More information is printed out now if there is a wrong result error in Spark job test

Bug fixes

  • BUGFIX: Fixed Connection reset error caused by HiveServer2 closing long-standing connections. This should also fix most of these error messages: The following Hadoop configuration properties must be defined, but a connection error prevented the automatic retrieval of these properties
  • BUGFIX: Default output compression for Spark jobs is now always disabled, which should solve Wrong result errors for Spark job test
  • BUGFIX: Using a non-Kerberized connection after a Kerberized connection no longer fails with the following error: Server asks us to fall back to SIMPLE auth, but this client is configured to only allow secure connections. Note that concurrent use of such connections is still not supported
  • BUGFIX: Submitted job in Spark job test is now always killed if there is a timeout error
  • BUGFIX: Typo fix: Hadoop version "Mapr 5.x" changed to "MapR 5.x"